|Date: Friday, October 20, 2017
Location: 1084 East Hall (4:10 PM to 5:00 PM)
Title: Compressed Sensing and Machine Learning in Single Cell RNA-sequencing
Abstract: Until very recently, biologists could only sequence RNA in bulk (each batch containing on the order of 10^3 to 10^7 cells). Thanks to breakthroughs in technology, we are now able to sequence RNA at the level of individual cells. Single cell RNA sequencing (scRNA-seq) gives biologists a unique insight into the heterogeneity of cells, even within the same tissue. The analysis of scRNA-seq data, however, poses algorithmic challenges due to the complexity of working with high-dimensional sparse data.
In this talk, I will discuss the machine learning problem of feature selection--in our context, selecting a small subset of genes that is most informative about the biological function of the cells. I will contrast a traditional approach based on statistics with methods based on mutual information and 1-bit compressed sensing. Time permitting, I will also share possible approaches to "imputing" data that we believe "dropped-out" as a result of single cell RNA-sequencing's shallow read depth (and explain what this means).
This talk will emphasize motivation and big picture connections to compressed sensing and machine learning and should be accessible to all grad students.
Speaker: Umang Varma
Institution: University of Michigan
Event Organizer: Audra McMillan email@example.com